A.J.Underwood AMIEE
March 1992 -
Radio Local Area Networks.
Subject :-
and solutions.
By A.J.Underwood AMIEE
For Newbridge Networks / SERC
From Swansea University.
Date :-
SERC Link project at Swansea University with Newbridge Networks Ltd.
A.J.Underwood AMIEE March
1992 -
Synopsis.
The principle thought behind this document is the design and development
of a high speed Radio Local Area Network system. The main goal is at least a 10M
bit medium transfer data rate including the data collisions that occur similar to
that of Ethernet. However as will be discovered, this has now become a problem of
the past, as the "Maximum Efficiency Network Interleaving Protocol" has now been
devised for the full complement of network users, designed also for the use of the
ultimate RLAN application, the ROAMING LAP-
The technical difficulties associated with the RLAN technology as currently seen is reviewed and primarily assessed, together with various way to establish the currently concerned RMS Delay Spread measurement. Computer models assessing the effects of various antenna types have been written, and on the surface it would seem that directional antennas are best as these reduce the radiation pattern of the antenna, with its associated reduced reception sensitivity to stray bouncing signals around the room, i.e. the multipath environment.
Various modulation modes are viewed and found that the best solution may to use synchronous demodulation techniques, due to their time domain selectivity to the multipath signal. However one must not forget the AM rejection ratio of frequency modulation, its capture effect upon FM signals and also the efficient use of the transmitter power supply particularly if one is roaming around on batteries. Here switching off the internal hard disc when not in use is an advantage, therefore relying upon the networks support.
However what if one is not able to reach an area controller, the solution
is simple, use someone else’s P.C. for a network radio repeater. There is one technology
that uses this sort of application and that is to speed up the AX25 PACKET RADIO
technology, inter-
SERC Link project at Swansea University with Newbridge Networks Ltd.
A.J.Underwood AMIEE March
1992 -
Index Page Number.
Chapter One.
Introduction 1
Introductory Radio LAN technical Discussion 2
Basic Mathematical System Model 2
Use of Spread Spectrum Technology 5
Hartley / Shannons Equation 7
R/C ratio of information encoding 8
RMS Delay Spread Counter Measures 9
Error Correction 11
Chapter Two.
Characterisation of the Radio Channel 12
RMS Delay Spread Values 12
Multiple Antenna Phasing / Antenna Diversity 14
RF Channel Hardware Environment 15
Chapter Three.
RMS Delay Spread CAD Model and Measurement 16
Wideband Channel Sounding 16
Multipath Analysis 17
Signal Vector Equalisation 18
CAD Propagation Study 20
Program Mythology 20
Multipath signal data study results 21
Motion Fading 22
Chapter Four.
Radio LAN operational distance ranging 23
Phase Shift Keying Methods 25
Chapter Five.
Digital Signal Processing to fight the multipath
envirnoment 28
SERC Link project at Swansea University with Newbridge Networks Ltd.
A.J.Underwood AMIEE March
1992 -
Code Orthogonal FDM / Bit Rate Reduction FDM 29
Encoder simulator program 29
Decoder simulator program 30
Technical implications 31
Chapter Six.
Technical Realisation for Radio interfacing into LAN's 33
Maximum Efficiency Network "Interleaving Protocol" 36
Network Set-
Radio Frequency Implementation 40
Technical Solutions supplied in IEE colloquium 42
Conclusions 43
References 46
Appendixs
47-
SERC Link project at Swansea University with Newbridge Networks Ltd.
10 Mbits per second Data Rate R.F. Link.
Chapter One.
Introduction
Radio local area networks represent a new bread of network servicing
for P.C's and to any associated data communications link. There are various design
variations on this theme, some using high gain antennas between buildings and then
also for inter-
Current commerical technology limits the data rate to around a 1Mbit/s link, but the goal is for higher data rates up to 10Mbit/s and baud rates beyond this are visualised. Current protocols are primerely intended for a cable or fibre medium, but are thought to be also suitable for the Radio (RF) environment. However some of these protocols are complex to administer, and thus there is room to be made for a simpler but more efficient protocol designed. Some cable or fibre data signals carry also the clocking information by using a "return to zero data" pulses. These unfortunately gives a sinx/x bandwidth that is far too great than the actual data baud rate in use. It is for this reason that the "none return to zero" data pulses really should only be used, as this half the original "return to zero" bandwidth. This is in itself is likely to surface within the unit cost price of the eventual product, as the softwares engineers time is also part of the cost equation.
The index outlines an extensive list of investigative areas, some of
which may seem to be basic but never-
The radio channel is characterised in terms of it RMS delay, and by considering a practical dicussion papers covering various types of factory and office environments, an overall blue print can be drawn up. Antenna diversity and phasing is also included and based on a discovery made by an error found within an authors paper, an extract of which is included.
Technical methods to measure the RMS delay spread are considered, while at this point the comparison to television signal ghosting becomes apparent. A 2D CAD model is used to determine the effects of an antenna gain on the RMS delay spread to the transmitted data pulse. Radio LAN operational RF budget is mathematically modelled to determine the minimum transmitter power, but the overall losses from the environment have not been considered. This indicates the necessity of RF transmitter requiring power control.
Following the recent trends, Digital Signal Processing is viewed, but for the data rate in use, this may most certianly prove too costly. A broadcasting method known as Code Orthogonal FDM or Bit Rate Reduction FDM to relay the RLAN data instead of Digital Audio Broadcasting. The principles are considered but the technology is far from the final chip set.
The complete project is then viewed in the "Technical Realisation" for radio interfacing into LAN's. Within this discussion the final overall future design is outlined and graphically illustrated, using the "AX25 Packet Radio" and the "Interleaving Telecom's Protocol".
Introductory Radio LAN technical Discussion.
As the radio technology solution is to be used for the ultimate portability,
it is wise to study the environment in which the radio link will be asked to survive.
A prelimary study towards the basic requirements of the RF link budget can be viewed
from a mathematical point of view. The various pit-
Basic Mathematical System Model.
In this section, we will concentrate on the various parameters are used to analysis the fundamental RLAN system performance. Various topical subjects associated with RLAN technology are viewed and sized up against the appropriate restrictions that govern there performance within the field of RLAN operations.
Cross connections have also been tied between basic communications theory and ideas to combat the most problematic complication of the whole RLAN field, RMS delay spread and the associated Power delay profile. Together they represent the decaying multipath signal for the transmitted radio data bit, of which both are best considered separately.
The multipath propagation environment produces a series of delayed signals
arriving at one point, the RMS time average of which is called the RMS delay spread,
"". As the bit period approaches to that of the RMS delay spread value, then the
delayed signal will over lap the intended signal, causing a near continuous series
of inter-
For digital transmissions, the RMS delay spread is often quoted normalised to the bit rate, "R", in a similar way to normalising impedance's to 50 ohms. The resulting RMS normalised delay spread "d", allows comparisons to be made between systems operating at different bit rates, equation (1).
d = .R equation (1)
This equation suggests an upper limit on the bit rate of the signal transmitted
via an radio channel, the limit depending primarily on the RMS delay spread for the
current radio channel. Various RMS delay spreads have been reported from about 25ns
for a medium sized office to 125ns in a large office buildings. On the other hand
these figures have been quoted to vary from 50 -
Upon investigation, the Inter-
F = 1/ T equation (2)
where T = time delay of the multipath signal.
This phenomena may be related to as an inverted "Coherence Channel", a channel bandwidth whereby a group of frequencies are effected by signal fading. The bandwidth of the coherence channel can be determined by the range of phase shift to a frequency span placed upon the intended signal. It must be made aware that the interfering effect of the multipath signal will also be dependent upon the signal strength of that particular multipath signal to affect the intended signal. If the propagating distances are similar but different in phase by a fraction of a wavelength, then a vector summation would be significant.
Computer modelling will allow one to appreciate the mechanics behind a multipath signal phenomena within an enclosed environment. One would then be in a position to provide a informed solution.
One of the major requirements of a Radio LAN Network is an adequate signal to noise ratio for the received signal, which intern limits the degree of signal fading and operating range. The Bit Error Rate is also affected by the signal to noise ratio of the received signal, the wider the transmission bandwidth the greater the required received signal strength in order to overcome the increased noise floor of the wider I.F. band pass, equation (3). Alternatively the transmitter power may be increased, but at the expense of stronger multipath signals. An increase RX sensitivity has the same effect as increasing the transmitter power.
Rx = (KTB (s/n)) + fading ratio Watts equation (3)
Where Nbw = noise power bandwidth ( as B increase so does Nbw )
T = absolute temperature, Kelvin's
k = Boltzmans constant, 1.38 * 10-
B = Bandwidth ( Data Rate * ½ ) Hz
Fading margin = magnitude of signal fading.
Various methods can be used to overcome the limiting effects of signal fading, either by a twin antenna system with a separation of a number of wavelengths, or by a use a modulation system that can maintain communications even when comfronted by extreme conditions.
Positioning an antenna system such as to minimise the multipath signal content has obvious advantages. One may use a dual antenna system to create an "Antenna Diversity " principle, but these systems place the antenna at wavelengths apart in order to choose the strongest intended signal. RLAN designers do not have the luxury of space to seperate two antenna wavelengths "" apart placed upon the RLAN interface. Antennas used for a communications link placed in order wavelengths apart, exhibit a combined radiation pattern, which in accordance to "line of sight" theory can be modelled to provide an insight into the antennas diversitys effectiveness is discussed
latter in this document.
The use of a modulation system that is best suited to the RLAN application
is most certainly food for a hot debate. One should bear in mind that the signal
to noise ratio performances change with each modulation mode, and this should be
accounted for when concerning adverse signalling conditions. On top of this, a M-
Use of Spread Spectrum Technology.
By merely pronouncing the phrase "Spread Spectrum Technology", one immediately cungers up visions of secured radio conversations, used for rather important communications. However this is by no means the technology of the few, as with today's highly integrated radio equipment, adding the necessary "Acquisition and Tracking circuitry" to the modified radio is in the end result essentially just a matter of adding another chip.
The spread code gives rise to the use of Code Division Multiple Access
(CDMA) communications. CDMA technology allows the use of multiple number of user
codes on a single channel allocation. The acceptance of the wanted CDMA code(s) (multiple
orthogonal codes, Ref 4) is the prerogative of the receiving radio data modem and
then the subsequent suppressing of the unwanted code(s) by the Anti-
The length of the spreading code of a Direct Sequence system is influenced by a number of factors. As the Process gain is increased, the number of spreading bits to a single information data bit also increases, consequently for a fixed spreading data rate limited by the RMS delay spread, this can only reduce the effective information data throughput. A balance must therefore be found between the spreading code rate and the information data rate to the maximum baud rate possible over the radio channel. The length of the spreading sequence does not come into the equation until one decides upon the number of spreading bits per data bit, i.e. 127 spreading bits to one data bit accumulating to a process again of 21dB. However, one may also use a number of sequence codes in a repeating fashion to one data bit. The Process Gain is now viewed as the ratio of two data rates, equation (4),
spreading rate
Pg dB = 10 log10 -
data rate
One must also bear in mind that the acquisition and tracking of the Spreading
code will take a finite time period to lock. By using alternate spreading codes per
data bit, the receiver will find this difficult to acheive unless some sort of pre-
Although the use of direct sequence spectrum spreading has been thought
as un-
Unfortunately this view is not holly correct as a comb transmitter will
be able to block the channels where as a direct sequence method will be able to spread
out and therefore suppress the interfering transmitter up to the anti-
Aj (dB) = Pg (dB) -
where Aj = anti jamming margin
Pg = Process gain
Il = Insertion loss
Time Division Multiple Access (TDMA) allows the sequential multiple use
of a single data channel and allows the en-
By adding Frequency Hopping to the communications link, the overall channel
characteristics will be randomised as the hopping sequence is deterministic in fashion.
However the frequency hopping process will not overcome the RMS Delay spread, as
the signal propagation delay time to the receiving point will be far shorter than
the settling time of the frequency synthesiser. Direct Digital Synthesis (DDS) technology
is far faster in the order of micro seconds, although a Prescaling synthesiser has
a settling time in the order of several tens of milli-
Hartley / Shannon's Equation.
One can easily realise that the effective transmission bandwidth cannot be allowed to go beyond the 3dB point of the I.F. bandpass filter, as this represents the highest frequency component within the signal and at the 3dB point, equal to the half power point of the carrier signal modulation bandwidth. Many techniques have been proposed to increase the maximum bit rate within a limited bandwidth, combating also the limiting effects of the RMS delay spread. In order to acheive this, these techniques have become progressively more complex.
It has been found that the RMS Delay Spread and Shannon's theorem are
inter-
d = [ Bw log2 (1+s/n) ] equation (6)
or alternatively,
d = [ Bw 3.32 log10 (1+s/n)] equation (6a), ref 11.
Shannon's equation depicts the maximum data rate of the communications channel in terms of the signal to noise ratio, bandwidth and channel capacity, indicating the requirement of greater signal to noise ratio for an increased channel capacity "d" for the same transmitter bandwidth.
The transmitter output power "watts" is contained within the equation
by virtue of the signal bandwidth "BW" and "s/n" ratio. The baseband noise component
will increase to Np = KTB, and this is the same for AM and FM. The measurement component
in bits/watts is merely the transmitter power spread over the bandwidth, but in two
very different ways. For AM the rise is linear over its bandwidth ( CW has a lower
noise floor than AM ), but for FM it is in accordance to the Bessel Functions. One
can pre-
Some of us may view Shannon's equation in another light, and that is the "Matched Filter" principle. Equation 6b, shows that the log10(1+s/n) is multplied by the factor 3.32, and that this multiplies the required bandwidth required for a set data rate by 3.32.
Bandwidth = Data rate 3.32 log10(1+s/n) equation 6b
This then allows a pulse rise time of 33%of its pulse length, but for a sinewave approximation then only the fundemental harmonic of the data pulse is required. Equation 6c, then illustrates the adjusted equation.
Bandwidth = Data rate log10(1+s/n) equation 6c
The RMS delay spread part of the equation(6) restricts the data rate to that of the normalised delay spread of the room. The combined equation therefore defines the four most important parameters of the data link, bandwidth, signal to noise ratio, channel capacity defining the RF part of the link and the normalised delay spread "d" defining the room parameters. One now therefore has a basic model of the RLAN interface to hand.
Minimum S/N ratio for Shannon's equation without code extensions.
Considering equation 6c, the matched filter condition is when the data rate bandwidth is equal to the transmission bandwidth. In this case,
1 = log10(1+s/n)
thus antilog10(10) = 1 + s/n
therefore s/n = 9
finally s/n = 9.5dB
This is the minimum value of s/n ratio that can be substainable by Shannon's equation without using error corrections codes to reduce the BER rate dependency to the noise background.
R/C ratio of information encoding.
A method in which the efficiency of an data encoding system can be referenced
is by the R/C ratio. This is a ratio of the information data rate "R" to the effective
channel capacity bandwidth "C", i.e. R/C = 8, indicating that there is 8 times
more information "R" passing through the channel capacity "C" than the transmission
bandwidth suggests. This method is termed as "M-
In order to the ratio of R/C >1, a multi-
Contained within Ref(10), 4 level (2+2) QPSK signal is used, equating
to a 4bit parallel data bus with 2bits/carrier, R/C = 4. The maximum frequency contained
within a data stream is twice the bit period, thus in this case a 10Mbit data link
is encoded into a 2.5Mbit data capacity link, which when modulated onto a carrier
signal has a maximum transmission bandwidth of 5MHz. Our 10Mbit 4QPSK system design
with a 5dB noise figure receiver, would thus require a signal input strength of -
Another method to determine the effectiveness of an data modulation method
in the face of the RMS delay spread, is the rather un-
Many serial data interface chips use a synchronising clock at 16 times
the data baud rate and samples within the mid point of the data bit period. This
corresponds to a normalised delay spread of 0.5, or a data bit slip of 50%. If we
use this as a maximum overlaping value, then for a 25ns RMS delay spread, this equates
to a data signal rate of 20Mbits and for an RMS delay spread of 100ns, this equates
to 5Mbits. However this condition only bears fruit for the first data bit, as after
this point the multipath signal will start to inter-
RMS Delay Spread Counter Measures.
In order to counter-
However if one wishes to transmitting basically as fast as one can, then one must switch off the carrier signal completely to allow all the multipath signals bounce around the room off the walls etc, to decay to the noise floor of the receiver or to a point where the multipath signal is not effective. Once this process has been accomplished, one may also view the absence of the carrier signal as a logic "0" data bit. Therefore by the appearance of a carrier signal, one may view this as a logic "1" data bit. The combining effects of the RMS delay spread and following power delay profile, are used to extend the overall data bit period. While allowing an equal time period for the logic "0" data bit, the maximum data bit is then basically twice the period of the RMS Delay spread plus a sufficient signal decay time for the multipath signal. This type of modulation is know as either Delta or Pulsed Amplitude Modulation. It is interesting to note that the Pulse Sounding method also allows the multipaths signals to decay before transmitting another sounding pulse, therefore the data transmission medium can be said to be a modulated derivative of the pulse sounding method. However a basic review of other modulation formats is now considered.
The multi-
Any M-
If one wishes to use frequency shift keying, no matter what form it comes
in, then to signal at the RMS delay spread rate is equivalent to the frequency hopping
at the same rate. The problem is within the frequency shift keying, FSK, demodulator,
in which a 100% multi path delay data bit signal will only merge the intended frequency
tone with the exponential decaying multipath signal of the prevoius tone. The result
is that the FSK demodulator will most probably view two simultaneous tones, instead
of single tone. If one wishes to change frequency and re-
It is worth noting that Phase shift keying is a form of frequency modulation,
with the sinewave cycle terminated in-
Error Correction.
As a matter of course error correction algorithms are applied to data
sequences in order to correct any data errors. However, the normalised RMS delay
spread barrier is rolled back, an allowable increase speed of data rate evolves.
This advantage can be utilised for an enhanced channel coding algorithm with a greater
error correction bit overhead, maintaining the original data rate but using the increased
overall data rate to implant the increasing number of error correction bits. This
can be loosely described as a Coding Gain in terms of the number of data information
bits to error correction bits. One type of coding sequence that will comply to this
sernario is block error coding, quoted for example as [7,4] relating to four information
bits to 7 bits in each code word so that each code word has three redundant bits.
This is also known as Reed-
Ref (9)
Chapter Two.
Characteristication of the Radio Channel. (Ref 5)
In this section, past works on the subject have been included to understand the RMS delay spread studies. An opion is put forward at various stages in order to branch out on a interium conclusion.
RMS Delay Spread Values.
Wide band multipath measurements at 1300MHz have been measured in a factory
environment to determine the range of RMS Delay spread. Values between 30 to 300ns
were discovered whereas the line of sight values in the order of 96ns and a value
of 105ns for a line of sight that included some path obstructions within it. The
worst case of RMS delay spread was a value of 300ns for a open plan metal working
factory environment. Delay spreads were not correlated with the transmitter -
When a radio path is lightly obstructed, the first observable pulse generally
has a larger amplitude as compared to multipath components arriving later in the
profile, hence the Power Delay Profile. The majority of multipath power arrives within
50 to 250ns after the first observable signal. For the case of heavily obstructed
paths, the first observable signal is generally weaker than components which arrive
25 -
Measuring the RMS Delay spread is a tricky process to master. In one
published paper (ref 6), it was quoted that the RMS delay spread at 850MHz was 270ns,
at 1.7GHz it was 150ns and at 4.0 GHz the measurement was 130ns, all measured at
one point, a distance of 0.3metres. However this is clearly wrong, as the different
times indicate that the propagation velocity of "c" has altered while the multi-
203
BW = -
Gain ratio
The end result is a lower RMS Delay spread figure. The paper also mentions as the frequency ratio increased, the delay spread over a particular path direction was less correlated, as the various RMS delay spread values didn't match, due to a function of the antenna effective wavelength for a standard sized antenna. However the path loss is also function of frequency using the inverse square law, equation (8).
D
dB = 20 log10 ( 4* PI * -
where :-
D = the distance to target
In another paper, Ref 7, it mentions that the RMS delay spread values
at 910MHz and 1.75GHz were found to be within 2ns of each other. This goes to provide
foundation to the conclusion that the RMS delay spread does not alter with frequency.
The assumption here is that a log-
Multiple Antenna Phasing / Antenna Diversity.
To combat signal fading in a dynamic environment, antenna diversity techniques
can be used. This involves the use of multi number of antennas placed at fractions
of or at numbers of wavelengths apart. By correct phasing the combined vector summation
can be swepped through 360 degrees of rotation to seek out the transmitted signal
and in the process focusing the effective radiated area of the antenna system onto
the transmitting source. Antenna focusing can be used to reduce the effective radiated
area, which will intern reduce the number of multi-
For an antenna seperation over a number of wavelengths, the phasor vector
addition may suppress the R.F. signal found over a largish area. Although the R.F.
signal may be treated as a ray of light, it is never-
RF Channel Hardware Environment.
There are gross physical differences between office buildings and typical factories. Building construction techniques, floor arrangements, building contents and placement of walls and other partitions, are all factors which greatly affect the signal propagation, differ considerably between an office building and a factory. Unlike buildings and houses, factories typically have very few internal partitions. Aisles are arranged in an orthogonal manner, an intersecting fashion and are flanked by metal machinery or inventory cabinats. Ceiling and walls in newer factories are made of ribbed steel and metal ceiling trusswork is used.
Such data is necessary for determining limits on data rate due to channel intersymbol interference caused by the multipath component and also provide insight as to the location and intensity of scatters within buildings. Factory or office wireless communications are important for systems that envisage providing very high data rates between mobile robots, automated machinery, personnel communications and remote terminals in factories and offices of the future. A voice analogue signal is far more resilant than that of a high speed data signal to multipath signals.
The placing of each remote receiving terminal can be the difference between a good quality BER signal or one that has unsatisfactory performance. Placing the receiver in line of sight with the transmitter normally leads to a good recipe for success, but in the face of multi path signals one could experience the opposite, failure. One normally assumes that signal fading in a room occurs naturally and that there are pockets or holes like gravitational black holes around the universe positioned within the room. However this is not the case, as it is the vector addition of the signals found at the antenna socket that leads to signal fading. To avoid the multipath propagation problem, one simple finds a position within a room that the direct line of signal exists, but where the multi path signal is suppressed. By pointing the antenna downwards and shielding the antenna from reflecting signals by small vertical walls such that only a direct line of sight is visible to the transmitter point, then the receiver would not view the complete multipath signal and could possibly avoid the RMS Delay spread phenomena. Alternatively the use of a high gain antenna could be used, as the the effective surface area of an isotropic antenna is folded over to point to a particular direction focusing the transmitter power thereby precluding the apparent gain of an antenna. However a high gain antenna will still be subject to viewing the multipath signal, although be it at a suppressed amount due to the acuteness of the antennas radiation pattern.
Chapter Three.
RMS Delay Spread CAD Model and Measurement.
Wide band Channel Sounding.
Ref 8.
There are various types of channel sounding methods of which are used
to characterise the Radio Channel. The methods vary from a multiple number of carrier
signals to using Pseudo Random Binary sequences to sub-
The multiple frequency system known as Tone Ranging (ref 3), functions on the receivers ability to measure the unique phase relationship of the tones, that depends upon the relative distance of the receiver. By monitoring the signal strength the fading characteristics of the channel may be noted. In addition to this if one frequency is used as a reference and the absolute phase relationship of the other tones is known, then the phase distortion of the channel may also be noted. The Coherence Channel spacing may also be verified.
There is a method by which a series of snap shots of the radio channel
may be taken, known as the "Periodic Pulse Sounding Method". The duration of the
pulse determines the minimum echo-
If white noise is applied to a linear system, such as the RF channel
and then correlated with a time delayed version of itself, then the impulse response
of the channel is found. The time delay can then be used to measure the multipath
signal RMS Delay spread. This however requires an accurate time delay network with
a minimum resolution in the order of 10ns or less. The time delay network may consist
of a CCD delay line or a variable co-
An alternative detection circuit is the use of Surface Acoustic Waveform filters to match the Pseudo Random sequence, producing a correlation pulse at the output. Again the associated time delays give a measure of the RMS delay spread and the channel characteristic.
Most of the above methods require large bandwidths in order to record
the results as the measurement is in process. The "Swept time delay cross correlation
method" time slips the transmitter and receiver codes in order to produce a repetitive
pulse output. The frequency bit rate between the two codes referenced to the transmitter
bit rate after correlation, can give a high resolution of distance measurement. A
10MHz code rate will give a 30metres sub-
The cost of each method varies drastically as a SAW filter is a very expensive item. A Pseudo Random sequence receiver at high bit rates can the complicated to construct, whereas the repetitive pulse method requires a RF pulse generator and a radio receiver plus a timing mechanism. This is perhaps the cheapest solution to the measuring of the RMS delay spread.
However the minimum RMS delay spread can be approximated by physically measuring the dimensions of the room. The laboratory in which I am currently working has a dimension of H=3m, L=7m and W=7m, and a maximum once around reflected trip time of 50ns, at 0.3m/ns. The RMS delay spread may then be said to be approximately equal 50ns.
Multi-
Many of the channel models use ray tracing methods to determine the RMS Delay Spread and signal strengths at various places within the local area. The signal strengths can be drawn like contour lines on a map. Two examples from two different papers is given to describe the channel itself simplisticaly, equations (9) and (10).
L
H(t) = Bn exp (jn) ( t -
n=1
where Bn = magnitude of the ray
n = phase of the ray
L = total number of ray impulse's
tn = time delay
And,
r(t) = ak exp(-
Where r(t) = magnitude of the ray
ak = real attenuation factor (inversesquare law).
exp( -
k = time delay of the multi path signal.
p(t -
To create a wideband model each of the mutli-
The attenuation of the radio signal has been modelled as relative to the inverse square law, which of course effects the multipath environment, and thus by the strict controlling of the transmitter power level, the effects of the multipath signal strength can be kept to a minimum, i.e. below the detecting threshold of the data demodulator. In areas of shadowing from the direct signal, the multipath signal would be the dominant signal. This is commonly found when searching for a television signal and ending up pointing the antenna away from the transmitter, or finding the best signal strength for the TV is in the centre of the room. However it must be realised that what ever the transmitter power is set at, the multipath signal strength will be relative to it. There is more on the propagation latter on, but for now it is worth viewing the multipath signal as "Signal Ghosting", which is normally found visable on television set.
The multipath equations are very useful as they hold within them the channel characteristics of each received pulse. From a successive pulse analysis, the power delay profile may be plotted so that the RMS delay time referenced to the transmitted pulse may be calculated. Incidentally, the GSM standard transmits a 26 pulse sequence to learn about its environment.
The "real attenuation factor" is a calculated path loss, which for a
wavelength of 23cm and considering a range of 10 wavelengths, comes to path loss
of 41dB. The "phase shift" is due to propagation of the signal, which is a function
of the time delay of the multi path signal, that can be equated to a signal phase
shift referenced to the original intended signal. Finally for our equation description,
the last term p(t -
If one re-
1/k In [ r(t) / ak ] = 2 f equation (11)
In order to determine the waveform shape of the power delay profile, all the received multipath signals require to be considered. Each delayed signal has its own time delay "td" and this may be considered as a spectral point within the frequency domain, 1/td = F. When all the multipath signals are considered, then the inverse fourier transform may be used to determine the time domain plot, the very item shown on an oscilloscope. However with equation (10) one can calculate the value of "f" and by putting in place the RMS Delay spread value which is a time delay function, the RMS fundamental frequency and hence the minimum receiving bandwidth of the power delay profile may be found.
Signal Vector Equalisation.
It has been suggested by members of this group that a form of equalisation signal vector may be used to remove the interfering signal from the demodulation stage. However this may not be a practical solution, as the cancelling vector would be required to predict the arrival of the multipath signal referenced to the intended signal. Although the RMS delay function determines the time arrival of the multipath signal, it is the time dispersal or spread of the RMS delay function that continuously under shadows the information signal. This phenomena is as continuous as the transmission time of the information signal and therefore will merge together to from a constant interference signal. If the multipath signal has to travel twice the distance as the direct line of sight route, then the multipath signal is approximately 6dB's down on the intended signal strength, although the total reflection from a reflecting surface is a function of the frequency. However this is equivalent to two transmitters transmitting on the same frequency allocation. It is worth bearing in mind that carrier signal demodulations performance to an interference 6dB less in signal strength to its LOS signal.
CAD Propagation Study.
There has been much effort by researchers to determine the propagation characteristics of the radio environment. Mostly the work has been concentrated upon methods to measure the RMS Delay spread, and also to clearly understand the temperment of the multipath signal. However most of not all the researchers have not produced papers of the RMS delay quanity to the type of antenna used, although the writen papers do mention the type of antenna used, but have failed to determine the changing effects due to the various radiation patterns of a 5/8 over an istropic or even a 3 element antenna.
It is clear that the change of RMS delay to frequency does not alter,
as the propagation velocity of the RF signal is constant over the radio spectrum,
i.e. 1metre per 0.3ns. One researcher quoted that the RMS delay did not change due
to frequency, but their figures where 2ns apart, accountable to the slight variation
of the radiation pattern from the log-
The fundemental problem with an enclosed environment RLAN network is
the power delay profile, i.e. the decaying signal lingering due to the multipath
environment. This can be modelled, as the length of the delayed profile is representative
of the overall surroundings, hence once again the RMS delay value. The effect of
this will be to possibly raise up the trailing edge of the data pulse. To study this
phenomena, the multipath data previously accumulated from a computer model for various
antenna positions within the room, is re-
Program Mithology.
The program ray traces the room until the last ray form the incremented radiation pattern comes upon its intended target. The travel time period of the multipath signal will be depended upon the angle it left the antenna. Thus if the radiation angle is reduced, the travel time will be less. This is essentially the time period that will eventually determine the transmission period between each data pulse, irrespective of the modulation format.
In order that the multipath phenomena may be studied, a Ray Tracing program has been written to characteristics the propagating R.F. signal. The inverse square law has been assumed to be a true reflection of the signal attenuation characteristic to the distance travelled. Various other approximations of signal attenuation to distance such as the inverse fourth power law, has not yet been taken into account, although duly considered. It has been found that within the mobile radio environment, the inverse fourth power loss curve most probably occurs after a number of wavelengths of the transmitted signal, Ref (13). Within the space of a enclosed environment, such as an office room the inverse square law has only been applied. To assess the shape of the induced signal into the antenna socket in a two dimensional form, a Delta pulse shapped signal is assumed. This is most apparent as each induced individual multipath signal will arrive at the receiving antenna with its own signal amplitude and phase. The induced signal will be the vector addition of the received signal. This problem is surmounted and easily solved. The end result from the program is the vector amplitude and phase of the induced E.M.F. into the antenna socket. This is what is naturally displayed on an oscilloscope.
These graphs are drawn using the polar radiation pattern viewed vertically
along its directional axis. The omli-
In order to vectorally add the multipath signal, one simply converts the distance attenuation, time and compass bearing into meaningful figures. The distance travelled related to the wavelength of the signal reveals the phase information at the intended target. The distance attenuation coefficient is transformed into a numerical ratio, and thus used as the relative signal strength. The now amplitude and phase values are transformed into complex numbers, and duly added with the current combined vector addition. An 180 phase shift will of course suppress the overall induced signal strength thereby creating a fade within the signal. The depth of the fade will depend upon the room characteristics. However the various pulse timings are generated by overlapping the decay profile of the room. Once the pulse at "t = 0", to "t = 25ns" has elapsed then the decay profile is added against the rise time profile. The decay profile is the rise profile substracted from the overall signal offset by the data pulse length.
Multipath signal data study results.
The RMS spread delay function may be viewed in two forms. Firstly the RMS time of the overall data time period including the multipath signals, or secondly which is more likely, the time period related to the RMS energy within the data pulse including the multipath signals. However the time delay function must only be viewed as the time period in which the pulse multipath sequence falls below the minimum detectable threshold of the data demodulator, which includes the vector addition of the multipath signal placed upon the new data bit from the last data bit pulse. In other words the overlapping of the data bits due to the multipath time slipping the intended data bit.
The room does not seem to posses a low pass filter approximation, but one of a high pass characteristic. This was found by the CAD analysis as the lingering effects are not so apparent to the pulse length, if pulsed less than the RMS Delay spread timing. It is seen that the 50ns & 100ns isotropic signals linger on, but the 25 & 50ns and possibility the 100ns 5/8, do behave accordingly to the inverse square law. The same time delay measurement values can be found if a delta signal (pulse transmission) was used. The filter approximation due to the I.F. stage filtering, would transform the impluse into a raised cosine pulse shape and thus combine the overall signal
By changing the antennas radiation pattern, has confirmed that the multipath signal environment can be controlled by the antenna design. The group delay arrival of the multipath signal is depended upon the antennas position within the room, referenced from the transmitters positioning. The total length of the group delay arrival of the multipath signal upto to point were the last signal "hits" its target, will give the total time period for the delay spread function. This then indicates that the RMS delay spread may be found by the CAD modelling of the propagating environment. It is easily apparent that the maximum time period of the data signal to "hit" its intended target is equal to the delay spread function. In our analysis the transmitter is assumed to be active up until the last multipath signal is present at the target area. After this point the transmitter is switched off and the decaying pulse analysed. The total time period of the transmitter activity is in this case equal to the "Time delay period" of the overall data pulse. The graph plots also apparently demonstrate that any signal level change will not be stable until the delay spread function has ran its course. The isotropic 50ns pulse lingers on, but a 25ns pulse decays according to theory. The 5/8 equivalent does not seem to posses a lingering problem, and is approximately 10dB further attenuated at the room delay profile time, than the isotropic timing.
By using various antenna styles then the multipath environment can be controlled as shown by equation (7) ref (13). By using the ray plotting program then this effects upon the time delay function of the room can be measured, appendix "B" the graph analysis.
If a QPSK signal system with amplitude was applied to give a M-
Motion Fading.
This is the loss of Bit Error Rate associated with the movement of a person causing a Doppler shift of the transmitted signal or the dynamic moving of a multipath and intended signal. The Doppler shift will be small compared to the signal bandwidth, but the moving multipath signal will disrupt the RMS delay spread measurement causing the local area around the person to develop a signal fading pattern. As the person moves so will the signal fading area. This phenomena will be only be seen at frequencies where the human body begins to deflect or absorb the RF signal rather than being transparent to it. This is of course only an hypothetical idea. The end result maybe a BER graph similar to a mobile radio channel for continuously active location.
Chapter Four
Radio LAN operational distance ranging.
In this section an outline of the systems performance is illustrated via the use of a CAD Model. The transmitters propagation range using the classical inverse square law theory is modelled. The receivers performance is placed onto the graph by scaling the sensitivity to the attenuation of the transmitted signal. By cross referencing onto a distance scale, the operating range of the system in a open plan environment is estimated.
The receivers threshold trigger may be also lowered to provide a "FADING MARGIN", to protect against the loss of signal from multipaths signal and signal absorption, which is particularaly apparent as the human body is 70% full of water. A 2.4GHz signal is absorbed by water, as this is the operating frequency of a Micro wave ovens. A micro oven resonates the water molecules within the body, but this requires a high power rating. For safety sake, only low powers of 2.4GHz or any other microwave frequency should be used, as quoted within the DTI specifications of no more than 100mW IRP (isotropic radiated power) and therefore equal to the ERP, Effective Radiated Power.
To scale the required receiver sensitivity of a 10Mbits 16+16QAM, we can consider the complete problem in terms of dB, as follows.
Boltzmans constant = -
Boltzmans dBm con = 30.0 dB
Bandwidth = 63.98 dB (2.5MHz Bw of a 10Mbit 16+16QAM)
AM Rx s/n ratio = 16.32 dB (10.3dB s/n per DAC level @ 1*10-
Rx noise temp. = 27.97 dB (627K, 5dB)
-
I/p Rx signal = -
100 metre range @
2.4GHz attenuation = 80.05 dB
-
Transmitter power = -
This equates to a signal strength of 6.5uV, or approximately an S6 signal. Within the serious field of Radio Communications, this signal strength is well above what is ruled as a sensitive receiver and is some circles this is classified as being "deaf as a post"!. However as one can see, the transmitter power is around 86uW, and therefore one has room for manoeuvre. A one milliwat source would add 9.29dB to the signal to noise ratio, but the best radio network policy is one where by the lingering effects of the RMS delay profile are quickly buried into the radio receivers noise floor.
Appendix "C", illustrates the operating range of the radio communications link, in terms of the detection signal strength and the associated signal to noise ratio, but also it illustrates the range flexibility of the system to extend its range by simply increasing the transmitter signal power. However one must be aware that the RMS delay spread, or the time for the return pulse, increases with distance. This will then only reduce the data rate for some terminals, but the maximum data rate of the system is usually governed by the maximum data rate of the slowest terminal data rate, within a network environment.
In order to establish the operating range of the RLAN network, one may also consider the effective RMS delay Spread per unit distance, although it is the return path time of the delay signal that limits the data rate performance. If a room is narrower in one direction than in another, then the longest distance will possess the greatest delay spread value, although by the time the reflected signal returns to the intended receiver and depending upon the ratio of the distances between the intended and multipath signal, the signal strength may have died sufficiently enough as to not to interfere with the indirect multipath data signal. However if this is not the case, then the overall operating range to a specific data rate, equation (1), will be restricted.
In order that the RMS delay spread to distance ratio may be overcome,
one may use a variable data rate modem. However the modem design is set to 10Mbits
and cannot be increased any further. If this done, then inter-
In order to try and combined the RMS Delay Spread effects to the bit error rate, the signal to noise ratio of the link is calculated. In order to change the signal to noise ratio, equation (6a) modified to a Matched Filter, then the effective data rate carrying capacity of the room limited by the RMS delay spread is calculated. This value is then put into place within equation (6a), with the modem data rate set to 16Mbit and the new signal to noise value is then found for the system. The bit error rate for a QPSK mode of data modulation is the calculated. The folowing equations may be used, but as was found out, the answers are not clear, and only really indicate the maximum bit rate to a variable RMS delay spread to the bit error rate. Unless one designs a truely variable bit rate modem, the equations are not really worth drawing.
equation (1) d = .C ; where "C" is taken as the effective data rate of the room.
equation (6a) C = R. log10(1 + s/n) ; where s/n is calculated, modified to a Matched Filter.
equation (12) Pe = e (-
Shift Keying Methods.
There are various derivatives of the standard shift keying methods but
initially it is PSK. This is basically two quadrature carriers that are amplitude
modulated. The amplitude modulation method chosen is DSB-
Any PSK system that has more than one data bit encoded as a symbol bit
is referred to as M-
The Bit Error Rate is therefore calculated on the amplitude modulated
carrier, brought down to a baseband signal for quantization. As the carrier is also
bi-
The quadrature coherent carrier injection oscillator signal will correlated
only one of the quadrature signals therefore demultiplexing the simultaneous multiplexed
quadrature carriers. If one was to multiply the locally injected signal with the
incoming quadrature carriers, then only the in-
The recovering carrier signals are composed from the original data sequence,
by multiplying the signal by an amount equal to the M-
1
= tan-
2*PI*F*R*C
= degrees.
An alternative method may be viewed, which entails the use of a reference
signal amplitude modulated at a low level. This is then removed by the carrier recovery
circuit and used as a reference signal within a phase locked loop circuit. The reference
signal will be a sub multiple of the in-
The carrier signal is then bandpass filtered and in some instances injected into a Injection locked Oscillator, in order to maintain some degree coherence should the input signal suddenly drop out of sight. A "Blocking Oscillator" could be used as the resonance oscillation is retained until the kick voltage drops to zero. Today usually a phased locked loop is used with a long locking time in order not to view the sudden signal strength drop. Once the carrier is recovered, it is then divided down until it is 4 times the carrier frequency, were two "D type Flip/Flops" can be used to provide a quadrature carrier signals equal to ¼ of the recovered reference frequency. The recovered data signal is then two cosine data outputs.
In order to determine the Bit Error Rate (BER) value, one must refer to the constellation diagram. In order that the individual phasor may be detected the signal to noise ratio must be sufficient in order to allow a relative degree of noise resolution placed upon the signal so that each phasor vector does not merge into each other. One must bear in mind that the recovered data sequence is encoded as relative voltages placed along the quadrature axis, and that the relative data sequence will be split along both axis. This then means that the A/D conversion process is evaluating a ramp function encoded with the data sequence, figure 2. Before quantization begins, the data sequence would be put through a matched filter equal to the data rate, which helps to suppress the imposing multipath signal and preserve the signal to noise ratio.
It has been discussed that the signal detection is achieved at the A/D conversion stage, and that the signal data symbol resolution is defined by the minimum signal to noise ratio. It therefore follows that the BER count is determined by the resolution of a single constellation point and not of any other form. The minimal discernible difference between each constellation point is directly related to the signal to noise ratio, equation (14).
Pe = ½ erfc s/n equation (14).
However we a relating to the BER of quadrature data symbols, and therefore the probability values mutually add.
Pt = Pe1 + Pe2 equation (15).
Thus the BER is equal to:-
Pe = erfc s/n equation (16).
This equation is valid for all quadrature data signal modulation systems, independent
upon the degree of M-
This then leads us onto the point of the signal to noise ratio of the
receiver other that the QPSK demodulation stage. This is the 1KHz tone signal to
noise measurement value which depicts the quality of the receiver. For instances,
if each constellation point signal to noise ratio is 10dB, then for a 4 level M-
To add M-
Chapter Five.
Digital Signal Processing to fight the multipath environment.
The most basic form of digital signal processing "DSP" is within the area of data bit error correction. However this is not strictly a DSP function, as DSP operations are base on the digital processing of analogue signals. As shown within the BRRFDM system, the data bit stream sequence is converted into a parallel data bus format of frequency division multiplexed carrier signals. The major advantage with a DSP function is the integration of complex analogue signal processing into VLSI technology. However in the case of the BRRFDM principle then only DSP technology is really suitable, as a large number of carries are used. However one must not loss sight of the analogue alternative and any disadvantage or advantages must be clearly understood of both methods.
An area in which a DSP function may benefit is the un-
However probably the most effective defence against the multipath signal
is the analogue use of synchronise demodulation, namely referred to as coherent demodulation.
According to mathematical theory, the signal output of a synchronise demodulator
is depended upon the exact phase matching of the two signals. If phase coherence
results, then the output will be a full signal component. However if the phase alignment
is 90 degrees or PI/2 out of phase, then there is a zero signal output. Obviously
varying the phase between zero to PI/2 will give a rising output signal component.
The amplitude of the multipath component will now be additively attenuated by the
coherent demodulator, by virtue of the phase difference between the multipath component
and the referencing local oscillator carrier signal. Quadrature Phase Shift Key modulation
requires the use of synchronise demodulation, as well as posing the advantage of
applying a M-
The alternative approach to using a M-
Code Orthogonal FDM or Bit Rate Reduction FDM.
The first approach in order to study the COFDM or BRRFDM data encoding process, is to reproduce the Fourier Transform encoding process, converting a serial data format to a BRRFDM output, the Discrete fourier transform was used. This is easily achieved by using Fouriers original "discrete fouriers transform technique" with a computer program. Form this the time domain signal which is subsequently put forward to the Amplitude Modulator can be analysed in order to determine some of the "encoders" more immediate properties.
It is appreciated that a Fast Fourier Transform used for this process
is done completely in hardware, a chip set dedicated to the FFT process. The British
Broadcasting Co-
However, the complete COFDM/BRRFDM process is demonstrated in software using a very simple BASIC language program. The principle was also tried on COMDISC's SPW CAD package, but the decoder requires more understanding of the FFT block.
Encoder simulator program.
The plots shown in this case are computer predictions of the DAC output within a Bit Rate Reduction FDM system. The computer program produces the plots from Fouriers Discrete Fourier Transform equation, but without the scale reductions that is normally apparent with such an equation (17), i.e.
V(t) = amp*sin(wt) + amp*sin(2*wt) + etc .... amp*sin(8*wt) equation (17)
The above equation uses the variable "amp" to depict whether the data bit is logic one or zero, if logic one then a value of 30 is placed as the "amp" value.
The computer program can be extended up to any number of carriers, depending upon the screen resolution if one wishes to plot the waveform. A data word of FF hex was used for every byte, which revealed that the start / finish points show a sharp rise of signal amplitude for block of encoded data. As the numbers of carriers are doubled, the sharp amplitude rise extends by an extra 6dBV, increasing the dynamic range of the AMPLITUDE MODULATOR. It was found through experimentation, that a dynamic range of 20dBV was required for 8 carriers, and 26dBV for 16 carriers up to 32dBV for 32 carriers. By projecting these figures, a 1024 carrier system for 1024Kbit/s system, would require in the region of a 60dBV dynamic range. However its was found that a data word other than "FF" per byte gave a reduced dynamic range output from the program, by sum 6dBV. From this it transpires that the music data which is continuously varying will avoid a continuous FF Hex data word., appendix B
The highest frequency of the FDM process is easily distinguishable and that the lowest frequency carrier is not evident until it is removed from the program output with the lowest then equal to the second frequency component. The output then shows a linear rise through the plot output, one linear rise per half cycle of the lowest ( 1st) frequency carrier component. The lowest time period between the spiked output response will give the sinx/x response density for each carrier component. Essentially this time period is also equal to the encoded bit period of the Bit Rate Reduction FDM process., rather than labelling the data encoding principle as Code Orthogonal FDM.
A principle by which the peek to peek amplitudes may be reduced is to use a logarithmic conversion process, a log to base 10 scaling. A 100dB range will this be converted into a numerical scaled range of 100:1. This is very useful and easily accomplished (used on a Spectrum Analyser I.F. stage), as it will serve to reduce to dynamic range required of the Amplitude Modulator, in order that the FDM data signal may be carried onto a R.F. carrier signal. The output stage of the transmitter will naturally be linear in nature.
Decoder simulator program.
For a brief experiment, a DFT as apposed to a Fast fourie equations for simplisity was used. The overall principle turns out to be rather straight forward, the complete demonstration program testing over 8 discrete carriers.
The decoder program initially came from "Electronics and Wireless Worlds" "Interfacing with C" book. Essentially the equation tunes itself to establish the spectral response contained within the sampled analogue waveform. This is achieved by the ratio of "2m n/N", where "n" is the sample number and "m" the test frequency, with "N" the sampling frequency. The ej function is the rotating vector to essentially synchronisely demodulate in mathematics the analogue data signal to a zero hertz I.F. frequency (a direct conversion receiver), equation (18).
N-
X(jw) = x(n) e j(2n m/N) equation (18)
n=0
The X(jw) value is an amplitude ( or signal strength) value of the test frequency
"2m" within the sample "n", which is then threshold detected, i.e. if X(jw) > 0.5 to determine the existance of a carrier for logic "1" or logic "0" for X(jw) < 0.5, as a numerical answered value to the equation. The maximum test frequency is 2N/2 radians/second.
The major trick with the calculations is this, if N = 8000Hz and therefore
m = 4000Hz equal to N/2, then the m/N ratio may be normalised to 4/8. Incrementing
"m" form 1 -
The fundemental problem with BRRFDM is the processing speed of the FFT function. If a 1Mbit data link is encoded into 1000 carriers, then the sampling frequency is 2MHz. The processing speed is then up to 2 million times greater than is required for a one bit/second program run time.
Technical implications.
The dynamic range of the multicarrier output needs to be carefully monitored, as the each part of the output signal contains a great deal of information. Within the radio receiver, AGC circuits will need to be used as the BRRFDM signal cannot process the signal as one does with FM, as it is the unhindered dynamic range of the signal that is very important as it contains the vector addition of the combined carriers and hence the data information within it. However the data encoding format does give an advantage of not being sensitive to D.C. fluctuations within the amplifying circuitry, as provided the signal is faithfully reproduced. It is a series of digital number variations within the ADC conversion process that contains the multicarrier information, from which the IDFT (Inverse Discrete Fourier Transform) program will extract the data bit information content.
The large dynamic range required to satisfy the modulation format, is in itself for data information is quite a stringent specification, as it is 20dBV more than is required for television at 40dBV s/n ratio. Having now to treating the signal as the equivalent to video information signal changes the ball game completely. The bandwidth for the Digital Audio Broadcast" is in the order of 3MHz, and upto 5 times that for the RLAN project data rate medium. It must be remembered that the transmission bandwidth is twicw the data baud rate, i.e. is AM for carrier modulation of the BRRFDM signal, at 1Mbit the TX BW is 2MHz, and for 10Mbit the TX BW is 20MHz.
Normally the ADC conversion process is completed at the Nyquist rate,
i.e. twice the highest frequency component of the information signal. However this
may not guarantee that the ADC sampling will co-
For serial data format interface chips, such as the RS232, the sampling period of the data sequence is at the mid point range, i.e. 50% along the bit period. It is thus possible to slip the time period of the reflected by upto 40% of the bit period. The COFDM or BRRFDM signalling format parallel ups the data sequence such that the parallel data rate is far less than the serial data rate. The DAB rate of 1.5Mbits/s compressed into 1Kbits/s by using 1500 carrier signals, one carrier per data bit, corresponds to a bit period of 1ms, or a total reflected path of 300KM. This type of modulation format has a quite stringent specification, as the signal to noise ratio required reviles that of television at 40dBV s/n ratio.
Having now to treat a data signal as the equivalent to video information signal changes the ball game completely. Any phase distortion as result of the I.F. stage filtering will cause a distortion within the complex time domain signal, but will however not be detected within the FFT transform signal amplitude bearing output provided the said carrier has not been amplitude distorted within the time domain, but if so the signal must still be above the detecting threshold for a logic one data bit. Apart form the introduction of bogus carriers from 3rd order intermodulation distortion products and impulse noises, the modulation format may well prove to be quite reasilant.
The up-
Chapter Six
Technical Realisation for Radio interfacing into LAN's.
The Department of Trade and Industry's Radio Communications Agency has issued a specification booklet for RLAN technology for spot frequencies around 2.4GHz, 10GHz and 24GHz. The most attractive frequency is the 2.4GHz band, as the cost of the R.F. components are modest compared to the higher bands. The wavelengths for the lower 2.4GHz band is easier to handle by comparison to the 10GHz band, but for the 24GHz band the problem manifests itself to a far greater extent. The lower 2.4GHz band, the production resolutions for manufacturing errors will be far more torrellent, although the production stills for the 10GHz band which is basically the microwave "K" band, have been well learnt and therefore the difficulties associated with this band will have already been overcome. This is most specially the case for the Direct Broadcasting Satellite TV / RADIO "Ku" band.
Two of the most greatest potential faults within a radio link for RLAN networks are both related to signal failure. This is amounts to signal fading through either failure or insufficient transmitter power or as a result of propagation phenomenalies. This in itself breaks down into two forms, both as a cause of multipath propagation, be it Rayleigh or Rican. As with the placing of a portable television within a room in order to find the best picture signal, or in this case a drop within the Bit Error Rate as one discovers pockets of high signal strength. This signal propagation phenomenally is known as "Rican". The signal reflections result in a time delayed signal, i.e. the multipath signal known within the television trade as "ghosting". It is this phenomenally of signal ghosting that is a cause of great concern within the RLAN technology, the merging of reflected data signals onto the intended signal upsetting the balance of the encoded data sequence contained within it.
The Band plan bandwidth allocation is approximately 100MHz, for which
the appropriate data modulation format will have to be chosen for the data rate is
use. For example, using MQPSK a 10Mbit data link can be comfined to 2.5MHz of bandwidth,
while a 100Mbit link will occupy 25MHz of bandwidth. This nessitates the use of 16+16
level QPSK, giving the equivalent of a 8bit parallel data bus. The technology for
this modulation format is well and truly established. The transmitter power output
would be viewed with the transmission bandwidth of the data signal under consideration
as the required transmitter power the bandwidth is proportional to one another. The
power/hertz philosophy of measurement gives a ratio at the detector end to determine
the degree of noise associated with a variable data rate signalling system, as another
way to combat the RMS delay spread. The modulation format as previously mentioned
nessitates the requirement of a specific design of transmitter output stage, and
therefore the efficiency of the overall radio interface. By forward thinking, one
may invisage the ultimate use of the RLAN Network, for the data linking of a romming
LAP-
The antenna design is one place were the effectiveness of the RLAN can
be made and broken. The wrong antenna design can seriously reduce the performance
of the portability of the data terminal equipment. One may be able to walk around
the plant and be able to interface to the data network form some very in-
An application of a local area radio data network within the perimeter of the firm could be the test and calibration of equipment, where by the technical information of the equipment would be required to assess the quality control of the manufacturing process. One underground scenario could be for equipment calibration down a man hole for example?, and external antenna could be handy in this case. Adding new machinery into the closed loop production cycle would now be just a matter of adding a radio link back to the control equipment related to that machine. There would not now be a need to lay out the costly expense of new control cabling. Any building that dates back as far as the second world war is not likely to walls greater than a one foot thick including the gavity spacing. Only the listed buildings that are costly to heat are likely to have walls two foot thick and made of stone and not brick. The excuse then of costly of wiring up of list buildings may well be few, but depending upon the implementation of the radio network, cabling will have to be used.
It would be wise to pitch the cost any Radio LAN interface board at a
small percentage of the complete hardware costs of a P.C., 10 -
There is however another side to this discussion, and that is that Ethernet
the most probably used data link system for small systems, has a data packet maximum
length of 1500bytes and that although the data bit rate is 4 or 16 Mbit, the user
capacity is at maximum 30%, due to data collisions. If the effective through put
data rate is raised significantly through an Radio interface, then one may be able
to justify the extra costing of the R.F. alternative over and above the £100.00 Ethernet
card. Essentially what this all boils down to is that one may effectively be able
to use a lower data rate standard but at the same time be able to meet the same effective
data rate through-
There is one current R.F. alternative that is currently being developed
to a production level and that is the DECT standard (Digital European Cordless Telecommunications)
chip set. It is arranged into 10 channels of simplex data, each channel containing
12 packets of 32Kbits/s at a 1Mbit rate. The practical implications of this system
is the following:-
Each channel contains 12 packets, @ 32Kbits, which equals 0.393216Mbits of information
per channel in one shot. There are ten parallel channels of data, thus the total
data rate into the DECT chip set to facilitate an all channel coherent data transfer
is:-
10 (parallel ch.) * 12 (32K packets) = 3.93216Mbits of information, equation (19)
The time taken for a one way ( i.e. simplex) communications link at a 4Mbit data continuous link would be one second. The effective one way data rate is equal to one data block size, basically 32Kbits/s, as specified by the information sheet for computer born data. In order that voice can be accommodated, the switching of the simplex data channel must be fast enough as that the burst data up dating is perceived inpreceptable to the voice user.
There is a view to interfacing the DECT network onto the Ethernet system, but as the multi user efficiency rating of Ethernet is 30% the max user capability, it would therefore be imperative that the DECT to Ethernet connection must be solely a dedicated single user connection, single user being the DECT application. If two DECT's are used then the effective data rate is 16Kbits/s, and so on. The cost of this technology was last priced at around £1000.00 per P.C. interface box, rather too costly. Clearly an alternatively cheaper method is really required.
The data interface method used could be a packet data type of protocol, i.e. for example based on British Telecom's X25 data protocol for easy access to the world wide data network. To design a system for a rate of 64Kbytes per second would be interesting, 16 times faster than DECT. This could be time synchronised to avoid data collision, although this is not so much of a problem as the squelch circuit would be used to hinder the data transmission of other users. If it is correctly designed, then the efficiency rating should reach 100%. In actual fact the idea put forward here is suitable for all types of primerely serial data applications. Using frequency modulation will give an efficient transmitter design structure.
The X25 protocol need not really go much higher than the second layer protocol of packet radio technology, which is the physical layer digital interface, and the SDLC data link layer for the HDLC and associated control software. In order to avoid the reduced efficiency problems with data collisions on the network, a data access "Interleaving Telecom's Protocol" is required, which is quite simply done.
Maximum Efficiency Network "Interleaving Telecom's Protocol".
The best performance of a multi-
The first trick to the problem is the use of a rotatory delay time cycle and by giving each radio interface is a unique number. Essentially this number is its queuing position, and each radio interface transmits its own number as its ident with the data sequence. Every other radio interface continuously monitors these numbers to determine where in the que they are. After the transmission has stopped that radio unit goes to the back of the que, and every radio interface knows that it is getting closer to the time when it to has to fire off. The maximum delay between two consecutive radio transmitters could several network clock cycles, known as the TX/RX transition period, the time taken by the radio interface station to switch from transmission to reception visualised through the network medium.
If the previous station number was No.102 and the next No.134, then No.134
delays period is 32 clock cycles (134-
Now the question arises to how the cyclic clock can be maintained if
no one radio interface has transmitted in order to time synchronise the cyclic delay
clocks of the radio interfaces. The answer is simple, if no transmission exists after
say 2.5 times the maximum delay period, say 256 clock cycles for 256 users, then
the area controller for that network is issue out a transmission pulse. This will
start all the clocks, and if there is not a transmission again for 2.5 times the
max delay, another transmission pulse will be issued. The reason for this is two
fold. One is so as not to allow any radio interface clock to go a stray, i.e. out
of synch with its neighbours, and secondly to provide a synchronising pulse to re-
If a reply is forth comming, then one can assume that the communication
has been successful. Although the system was first though of as for a radio network,
the protocol will work equally well for cable or fibre optic network. The data packet
circulates around the network once as a BUS typology is used, which is the cable
equivalent for a purely radio interface, with the co-
If the customer for who the data is for has not responded within a set time period, then it may be retransmitted in sequential order until say an event number has elapsed. In practice it is recommended that each station is left on a listen / monitoring mode, and any correctly addressed packets received are stored to the hard disc, i.e. in a "mail box" mode. It would be quiet possible to use someone else P.C. as a data reference library, interesting ?, as is so with Amateur Radio, or via a telephone mail box, i.e. the "Birmingham Bulletin Board" is just one such example. The Amateur Radio X25 protocol, AX25, can and does uses other AX25 modems as data repeaters which may access into other network systems by simply quoting a network interface repeater address.
As each radio interface has its own control program as part of the AX25 protocol plus the sequential control of the networks "Interleaving Protocol", then the network can look after itself quite easily look after itself. All that is needed as an extra, is a "missing pulse detector" to determine if the network requires a synchronising pulse to maintain the networks self clocking in order to run successfully the networks "Interleaving Protocol".
Distributing the data amongst the network will give a form of distributed data security as if one P.C. is "computer hacked", then only a smallish section of the networks data capacity could only be tapped into. The P.C. then can be used to deni access to the result of the network be quite literally a physical key. Only authorised users with physical keys will have access to the distributed data network. However scrambled data by byte interweaving based on a PRBS code within the RAM DISC can be used to secure the radio link to an outside access, as the coding algorithm will only be known by the network users P.C.'s.
Network Set-
The internal firms network would essentially run at the speed required
for a sufficient data baud rate for the application in use, but be at a rate that
is a multiple of 64Kbits/s could well be advantageous. This rate co-
An "area R.F. interface" could be used to act as the baud data rate converter,
demultiplexing and re-
A smaller network of 256 users would be a network data rate of 16.777216Mbits/s at a local rate of 32 users would be 2.097152Mbits/s at a through put rate of 8Kbytes/s maximum.
Todays chip technology would make this application arrive basically tomorrow,
in the time taken for the various committees to decide upon the specifications for
a high data rate RLAN network. The physical implementation of the radio interface
would be a radio transceiver quite literally down loading to or up loading data into
an RLAN RAM CARD, whereby the host computer would interrogate the memory store as
if it were RAM DISC. The data transfer block could be set the size of the RAM DISC,
but a maximum size of say 64Kbytes could be used. Memory is cheap these days at around
£50 per 1Mbyte @ 70ns so using computer memory to down load data is not a excuse
to throw the idea into the out tray under the heading of a "costly exercise". In
fact if anything using RAM DISC is an advantage as it allows the radio interface
to be totally transparent to the host computer. One should bear in mind that P.C.'s
have their own high capacity memory store, commonly known as the HARD DISC, thus
any network interface would be used to transfer just general traffic too and from
different persons, or the periodic test results from a closed loop manufacturing
production cycle. If the chief engineer is constantly on the move, then a distributed
packeted data radio network interface system might well be the answer connect to
his/her lap-
The portability given by an Radio interface may just be for walking around
the office form meeting to meeting interfacing the lap-
And now here is a WARNING. It might well be best to realise that this type of technology is well established, i.e. X25 over radio, i.e. AX25. The cost of such a product should therefore not be greater than £300.00 per P.C. card if manufactured correctly.
Radio Frequency Implementation.
An acrimin for this technology can be named as "Packet Radio Interleaving Telecom's", PRIT for short. The system is exactly as it sounds, interweaving packet of data information, but using the X25 protocol, with the ability to repeat through another users Radio Interface for wide area radio communications, if one is not able to directly access the network via an area interface. The AX25 Amateur Radio packet radio technology does just this, repeats through some ones elses packet radio repeater.
The R.F. interface design appendix "E", is based on a standard technology with due regard to the current standing of R.F. chip production. The design itself straight forward with one or two unusual properties. Figure 4, illustrates the design principle. The operating frequency is within the 2.4GHz band and may therefore use Plessey's standard suite of synthesiser chips. An oscillator centred at 800MHz is mixed with the synthesiser to produce the I.F. offset within the fundemental frequency source. The synthesiser does not have time to switch and settle while the data tranceiver turns form transmit to receive mode. The best way to acheive this is to constantly generate both the transmission carrier and the local oscillator together. By constantly generating the local oscillator, the receiver centre frequency will follow the transmitters operating frequency. To create a duplex operation, the 800MHz oscillator is changed for a different centre frequency. By virtue of the automatic frequency tuning of the receiver to the transmitter, the FSK modulation will not be visible to the FM demodulator, as the frequency shift will be transfered to the receiver and subsequently cancelling the FSK modulation into a constant carrier, creating then a transmission detect signal.
Direct frequency modulation has been chosen for simplicity, but more importantly the overall cost is kept to a minimum. Essentailly narrow band FM will be created to the information data rate, but by shaping the data signal into its fundemental frequency component, the spectral shape of the transmission signal can be controlled. The most ecomonic solution is a simple low pass filter configuration, tuned to the maximum frequency of the data signal, i.e. F Hz = baud rate / 2. This is known in the trade as "Tamed FM", of which there are many versions there of, Gaussian and Minimum shift FSK for example.
The receiver itself uses a Plessey FM demodulator which uses "threshold
extension" techniques for Direct Broadcast Satellite Television. The transition response
of this technique is not known, although it is purely the re-
The physical installation of the antenna system can be accomplished in
three ways. One is to use a "leaky feeder" principle, while another is to use seperate
cables from the central point leading to the controller, but this then deludes the
operation of the network architecture. The third and probably best solution is the
use of a co-
Depending upon the network set-
This can be achieved by leaving a few gaps within the data transmission
around the loop. Essentially if the network is informed that your are romming, then
it can deliberately leave room on the network medium with a specific length of frame,
to enable the romming P.C. to remote access into a another P.C. for repeater access
operation. By leaving room, the any P.C. will be able to receive the romming lap-
Technical Solutions supplied in IEE colloquium (Ref 2).
Two examples are discussed to illustrate how two modulation formats can be used to provide a high speed data Radio LAN. The technical solution are provided by an IEE Colloquium on Radio Local Area Networks, RLAN's.
An unaided 16 APSK system can tolerate a normalised delay spread of 0.21. With frequency hopping the value increases to 0.51. Once again, diversity improves the situation with values up to 0.68 and 0.72 becoming possible for switched and equal gain combining antenna methods. When considering our initial design specification figure of 100ns RMS delay spread, this then equates to a current maximum data rate of 7.2Mbits/s.
For a further example, the DECT specification states a data rate of 1.152Mbit/s
using GMSK. However by using QPSK this can be extended to 3.8Mbits/s for a normalised
delay spread of 0.38 (RMS delay spread of 100ns). By implementing an equal gain combining
method the data rate can be extended to 5.5Mbits/s (normalised RMS delay of 0.55).
By furthering the development to include frequency hopping, data rates between 8.1
-
Conclusion
One may view Radio Local Area Networks (RLAN) technology in two ways. One to respectively change from the cabled environment while the other view points is that what can be achieved by cable is satifactory and why therefore should we change. However the requirements for RLAN is no doubt creeping up from behind the horizon, and one does not need a telescope to view the potentials for this technology.
As ones P.C. becomes for ever increasingly smaller and more highly specified,
it becomes a desk companion and not the boss. Currently firms purchase bulky P.C.'s,
but one has the choice of a LAP-
The major question is how do we go about designing a network of this calibra. The data base is the matter of the companies involved, but it is generally agreed that the data network medium is best improved upon. If a 10Mbit Ethernet can only support around 250 combined users due to data collision, out of potentially more than 6000 address, then something needs to be improved upon. The Ethernet data packet length can be for one seriously extended, or made variable say up to a maximum of 8Kbytes or depending upon the memory capacity of the RLAN P.C. card. However what ever arises afterwards, must be efficient and cost effective.
The simplest way to introduce a cost effective system is to use an already
used standard. Ethernets packets are too small, although this is to reduce to error/second
within the Ethernet medium. By raising the data rate, the errors/second is higher,
but the overall data through-
When one refers to an synchronised or an orderly fashioned data transfer, one turns towards a thought of a sequential request to all users to if they wish to send the data. Actually although this view pint is valid, it is not however now the case.
Data transfer between various machines is achieved through the packet
addressing of each communications and this is essential if a multiple number of users
have been accessed to basically the same piece of wire. However as a radio link will
be eventually restrictive depending upon the R.F. terrian, then it may well be advisable
to use a small cell structure of R.F. network users. However this does not restrict
access or the port-
The interlaeving protocol allows the synchronise access to the data medium,
while the whole network tunes itself to the data loading of the network medium. However
there is not the requirement of a log-
The local area splitting allows the reduction of the RLAN interface data rate in order to accommodate the multipath propagation environment. The area interface will multiplex the data again in an orderly fashion at a higher data onto the network central medium, i.e. fibre optic cable. The cost savings to the RLAN P.C. interfaces are essentailly viewed in the bandwidth, transmitter R.F. power and chip complexity, i.e. various data rates and power consumptions. As each user P.C. is allocated an ident number, then one may access any area interface, but the delay time for this method to the central medium will be longer. However if the user finds his/herself unable to access an area interface, then they may repeat through a P.C. acting as the data repeat, but also with direct access for itself. The area controllers are basically high speed version, of the RLAN P.C. interface card acting in the repeating mode.
The communications standard that allows this data repeat feed through
principle is that of the AX25 Amateur Radio Packet Radio Protocol. This is linked
together with the Interleaving protocol, known as the Packet Radio Interleaving Telecoms,
PRIT. An example to re-
The combination of the above should provide a user friendly protable
data linked working environment, with a network access medium protocol of a distinct
nature, the Interleaving Telecom's Protocol. Combined with the Packet Radio AX25
technology, labelled now as " PACKET RADIO INTERLEAVING TELECOM's ( PRIT ) ", one
has a variable access to a data network of un-
The antenna system is one area when the communications link can be made
and broken. For a confined room, a ring antenna would be suitable, but for longer
distances, a directional antenna application would be required. The reason behind
these thoughts are two fold. One is the required radiation pattern and the other
is the unfortunate prominance of the multipath signal. The restrictive radiation
pattern of a directional antenna will reduce the strenghts of the multi-
The modulation format chosen for its simplisticity is frequency modulation. The AM rejection ratio will help to suppress the multipath delay signals, and will be especially effective with a beam antenna to concentrate the transmitted signal in one direction, while the surrounding signals transmitted or received from the antenna are attenuated by the antennas radiation pattern, thereby decreasing still further any multipath signals, the said named RMS Delay Spread.
References.
Ref 1:-
Ref 2:-
Ref 3:-
Ref 4:-
Ref 5:-
Ref 6:-
Ref 7:-
Ref 8:-
Ref 9:-
Ref 10:-
Ref 11:-
Ref 12:-
Ref 13:-
Ref 14:-
